not much happened today

**Liquid AI** held a launch event introducing new foundation models. **Anthropic** shared follow-up research on social bias and feature steering with their "Golden Gate Claude" feature. **Cohere** released multimodal Embed 3 embeddings models following Aya Expanse. There was misinformation about **GPT-5/Orion** debunked by **Sam Altman**. **Meta AI FAIR** announced **Open Materials 2024** with new models and datasets for inorganic materials discovery using the EquiformerV2 architecture. **Anthropic AI** demonstrated feature steering to balance social bias and model capabilities. **NVIDIA**'s **Llama-3.1-Nemotron-70B** ranked highly on the Arena leaderboard with style control. **Perplexity AI** expanded to 100M weekly queries with new finance and reasoning modes. **LangChain** emphasized real application integration with interactive frame interpolation. **Kestra** highlighted scalable event-driven workflows with open-source YAML-based orchestration. **OpenFLUX** optimized inference speed by doubling it through guidance LoRA training. Discussions on AI safety included trust dynamics between humans and AI, economic impacts of AI automation, and the White House AI National Security memo addressing cyber and biological risks. **LlamaIndex** showcased knowledge-backed agents for enhanced AI applications.